TESTING THE MOLECULAR CLOCK USING SIMULATED TREES, FOSSIL, AND MOLECULAR DATA

Given the vagaries of the fossil record, molecular clock methodology represents the only viable means of establishing an accurate evolutionary timescale. However, the molecular clock remains reliant on the fossil record to constrain the rate of molecular evolution. Although, calibration is a critical step in any molecular clock analysis it is often the least considered. Approaches to constraining evolutionary time differ in their use of biological and geological evidence, vary in complexity, and ultimately yield different divergence time estimates.

In the absence of the known evolutionary timescale, which – if any – of the available methods are able to estimate divergence times with accuracy or precision? In reality this can never be known because we can never know the true time of divergence. The solution is to use simulated data, where the relationship between fossil evidence, rate variation and genetic divergence is known. Methods for generating phylogenetic trees, molecular and fossil occurrence data are readily available, but have never before been combined to assess the veracity of the molecular clock. We achieve this aim using simulated fossil data based on models of random and non-random fossil recovery potential, speciation and extinction processes, and molecular evolution.

Our results demonstrate that when the distribution of fossils is random, existing methods perform extremely well. In particular, when confidence intervals based on fossil occurrence data are used to constrain the clock, divergence time estimates show consistent accuracy. However, when the distribution of fossils in non-random, while confidence intervals based on fossil occurrence data remain reasonably accurate, molecular estimates show a substantial decline in accuracy. Given that the real distribution of fossils is highly non-random, this represents a formidable challenge to accurate estimates of evolutionary rates and dates.